Unbound

[Arthur Chance]

On 01/12/2020 16:35, RW via freebsd-questions wrote:
> On Tue, 1 Dec 2020 12:42:07 +0000
> Arthur Chance wrote:
> 
> 
>>> I'd be interested to know whether it does work because someone once
>>> mentioned that local_unbound only supports forwarding, but I've
>>> never found anything to back that up.  
>>
>> From memory, if you've not set anything up the first time
>> local_unbound runs it reads your /etc/resolv.conf to gather your
>> resolvers and creates an unbound.conf that uses them for forwarding
>> for all domains. 
> 
> It set itself-up with google and opendns caches when I tried it.

That would have been taken from your resolv.conf

>> If you've set things up yourself it just runs with
>> that, but take a look at /etc/rc.d/local_unbound to find out which
>> file(s) it needs to see in order to avoid the default configuration
>> process.
> 
> It's all gets tied-in to dnsconf and the various configuration files
> come with warnings that they are autogenerated and could be
> overwritten. I have no need for dnsconf, so I just installed the
> package version which works recursively by default.
> 
> The question was more about whether the binary supports recursion.
> As I said, someone on the list (from the project) said that it doesn't.
> He may just have meant that they made it awkward to turn-off
> forwarding, but there's clearly something built-out of the local
> version:

It's always better to take a look at the source yourself. If you look at
the shell script

/usr/sbin/local-unbound-setup

which is what does the autoconfig if you don't have a predefined
unbound.conf, you'll see it can create either a pure forwarding config
or a recursive one if you give it "none" as the forwarders list.

The major differences between local-unbound and the unbound package are
a) the package is updated between releases and b) the package uses a
full kqueue based libevent rather than a select based internal libevent
that isn't as efficient under heavy load. Both are equally capable of
recursive resolution.

-- 
The number of people predicting the demise of Moore's Law doubles
every 18 months.